Accurate regulation and control method and system for anesthesia medicine pushing rate of electronic injection pump

Abstract

The invention relates to the technical field of anesthetic infusion, in particular to an accurate control method and system for the anesthetic pushing rate of an electronic injection pump, and the system comprises a data acquisition module, a central control unit, an execution layer driving device and user interaction terminal equipment. Through multi-dimensional data acquisition and a deep learning prediction model, in combination with individual characteristics and behavior modes of patients, precise prediction and dynamic regulation and control of anesthesia requirements are realized. The execution layer driving device adjusts the flow rate and dosage of the medicine based on the regulation and control instruction, and anesthesia balance of each patient is ensured. Responsiveness and adaptability of the anesthesia system can be improved, precise anesthesia is achieved, and operation safety and patient recovery quality are guaranteed.

Description

Technical Field

[0001] This invention belongs to the field of medical device technology, specifically a method and system for precise control of the anesthetic drug delivery rate of an electronic infusion pump. Background Technology

[0002] With the continuous development of medical technology and the increasing demand for surgery, the precise infusion of anesthetic drugs plays a crucial role in clinical surgery. However, existing anesthetic drug delivery devices often suffer from inaccurate infusion rates when dealing with individual patient differences and changes in surgical progress. This leads to fluctuations in anesthetic effects or deviations in drug dosage, affecting surgical safety and patient recovery quality. Taking electronic infusion pumps as an example, patient weight, age, metabolic rate, and surgical type all significantly influence the required amount of anesthetic drugs. If existing anesthetic drug delivery systems cannot adjust the infusion rate in real time, they may struggle to meet the dynamic control requirements for depth of anesthesia during surgery. Summary of the Invention

[0003] This invention relates to the field of anesthetic drug infusion technology, specifically to a method and system for precise control of the anesthetic drug delivery rate using an electronic infusion pump.

[0004] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: a method and system for precise control of the anesthetic drug delivery rate of an electronic infusion pump, comprising a data acquisition module, a central control unit, an execution layer drive device, and a user interaction terminal device.

[0005] The data acquisition module is distributed at different monitoring points in the operating room environment and on the patient's body. The data acquisition module includes physiological parameter sensors, drug flow meters, and surgical progress monitors, which are used to collect multidimensional data related to anesthetic drug infusion in real time. The multidimensional data includes the patient's heart rate, blood pressure, blood oxygen saturation, current drug flow rate, and surgical operation status. Each sensor and monitor interacts with the central control unit through wired or wireless communication.

[0006] The central control unit receives multidimensional data from the data acquisition module, processes the multidimensional data based on a preset data analysis algorithm, identifies and predicts changes in the anesthesia needs of different individual patients, obtains integrated prediction results, and generates control instructions based on the integrated prediction results.

[0007] The execution layer drive device includes a precision stepper motor and a flow regulating valve distributed inside the electronic injection pump. The execution layer drive device is communicatively connected to the central control unit, receives control commands from the central control unit, and dynamically adjusts the infusion rate and dosage of anesthetic drugs to meet the anesthesia needs of different patients.

[0008] The user interaction terminal device interacts with the central control unit, allowing medical staff to view the real-time anesthesia status and set anesthesia parameters for individual patients.

[0009] Preferably, the central control unit receives multidimensional data and processes the collected multidimensional data based on a preset data analysis algorithm to identify the differences in anesthesia needs of different individual patients and obtain the anesthesia needs of different patients.

[0010] The data analysis algorithm includes a deep learning-based anesthesia demand prediction model, which is used to predict the anesthesia demand of different patients in the future based on historical physiological data, surgical type and external environmental variables.

[0011] Furthermore, by combining individual patient characteristic data and predicting anesthesia needs based on patient behavior patterns, integrated prediction results are obtained. Anesthesia parameters are then adjusted based on these integrated prediction results, thereby more accurately reflecting the actual anesthesia needs of each patient and further improving the responsiveness and adaptability of the anesthesia system.

[0012] Preferably, the process by which the anesthesia demand prediction model predicts the anesthesia needs of different patients over a future period is as follows: Historical physiological data is acquired from physiological parameter sensors, drug flow meters, and surgical progress monitors deployed in the operating room environment, including heart rate, blood pressure, blood oxygen saturation, drug flow rate, and surgical operation status at different time periods; the historical physiological data is then classified and cleaned to remove outliers and erroneous data, ensuring the accuracy and completeness of the data; the organized historical physiological data is stored according to individual patients and time series, forming a historical physiological dataset that can be used for analysis and model training; the time series includes seconds, minutes, and hours.

[0013] The data receiving unit of the central control unit obtains surgical plans and patient medical records for a future period from the hospital information system, including predicted surgical types and durations. Kalman filtering is then used to remove short-term fluctuations and noise from the data, resulting in a surgical plan dataset. Data validation rules are applied to check the data's rationality, ensuring it meets the requirements for subsequent analysis and model input. Simultaneously, the data is converted to a format matching historical physiological data for unified processing, including adjusting the time format to a timestamp format consistent with historical data.

[0014] Features related to anesthesia requirements are extracted from historical physiological datasets and surgical planning datasets. These features are then normalized and mapped to [0, 1] to obtain historical physiological data features and surgical planning data features, respectively. This ensures that different features have the same dimension, avoiding the impact of large differences in feature values ​​on model training.

[0015] A convolutional neural network model was selected as the basic architecture for the anesthesia demand prediction model. The number of convolutional layers, the number of filters per layer, and the activation function were determined. This included setting up a neural network with four convolutional layers, with the number of filters per layer being [128, 64, 32, 16], and the activation function being the Leaky ReLU modified linear unit function. Through rules of thumb and multiple experiments, a model structure that could achieve good performance on the training set without overfitting was selected.

[0016] The processed historical physiological data features and corresponding historical anesthesia demand data are used as training samples and input into the constructed model for training. The Adam optimization algorithm is used to optimize and adjust the model parameters so that the model learns the mapping relationship between physiological data and anesthesia demand. The historical anesthesia demand data is obtained through actual anesthesia depth or feedback from medical staff.

[0017] During training, the dataset is divided into training, validation, and test sets in a ratio of 70%:15%:15%. The model's performance metrics, Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE), are evaluated on the validation set to monitor for overfitting or underfitting. If the performance metrics on the validation set no longer improve or begin to deteriorate, training is stopped, resulting in an anesthesia demand prediction model. This prevents the model from overlearning noise in the training data and losing its generalization ability.

[0018] The surgical plan data features are input into the anesthesia demand prediction model; the anesthesia demand prediction model calculates and outputs prediction results based on the learned patterns and rules, including the predicted anesthesia demand values ​​for different patients in the corresponding future time periods.

[0019] Preferably, anesthesia demand prediction is performed based on patient behavior patterns, incorporating individual patient characteristic data. The process is as follows: Individual characteristic information actively input by the patient is collected through a user interaction terminal device. This information includes the patient's age, weight, metabolic rate, allergy history, and whether there is a specific pattern of anesthesia tolerance, including sensitivity to certain drugs. Indirect data generated from the patient's interaction with the anesthesia system is collected and analyzed to extract individual characteristic data, including the frequency, magnitude, and timing of physiological responses during pre- and post-operative recovery. This data reflects the patient's real-time changes in anesthetic drug demand. Potential individual characteristic data is mined from the patient's past medical records, including peak and trough periods of anesthesia dosage during each surgery and trends in the total anesthesia volume for different surgical types, providing insight into the patient's anesthesia behavior patterns from the perspective of anesthesia usage. The collected individual patient characteristic data is cleaned to remove invalid, erroneous, or duplicate data records. Significantly unreasonable physiological response values ​​are corrected or deleted to ensure data accuracy and reliability. Data of different formats and types is converted to a unified format suitable for analysis. Relevant features are extracted based on the characteristics of the individual patient characteristic data and the prediction results. Composite features are constructed by combining physiological data and time information.

[0020] Cluster analysis is used to group patients with similar individual characteristics into the same category. A typical behavioral pattern model is established for each cluster. Association rule mining algorithm is used to discover the anesthesia response patterns of patients at different time points, analyze the frequency and conditions of their occurrence, and clarify the dynamic anesthesia demand change pattern of patients under different conditions.

[0021] By using statistical analysis methods, the average anesthesia demand and the range of fluctuation in anesthesia demand under different behavioral patterns are calculated. A mapping model between behavioral patterns and anesthesia demand is established to clarify the quantitative relationship between different patient behavioral patterns and anesthesia demand.

[0022] Simultaneously considering the influence of environmental factors on the relationship between patient behavior patterns and anesthesia needs, a comprehensive model was established using multiple regression analysis, which included surgical type, surgical duration, patient behavior pattern characteristics, and anesthesia needs among environmental variables. This model analyzed how patient behavior patterns affected changes in anesthesia needs under different environmental conditions, and obtained anesthesia needs prediction results based on patient behavior patterns, thereby more accurately predicting anesthesia needs.

[0023] A strategy was determined to integrate the anesthesia demand prediction results based on patient behavior patterns with the prediction models based on historical physiological data and surgical planning data. The integration strategies include: weighted average method: assigning weights according to the accuracy and importance of different models; stacked ensemble learning: using the prediction results of one model as input features of another model for secondary prediction.

[0024] Evaluate the performance of different fusion strategies on the training dataset, select the fusion strategy that optimizes the overall anesthesia prediction model, including the one with the smallest prediction error and the highest stability, and determine the final fusion strategy.

[0025] The anesthesia demand prediction results obtained from patient behavior pattern analysis are integrated with the output of the anesthesia demand prediction model based on historical physiological data and surgical plan data according to the final fusion strategy to obtain the integrated anesthesia demand prediction model and obtain the final anesthesia demand prediction value.

[0026] An integrated anesthesia demand prediction model is used to predict the anesthesia needs of different patients over a future period, resulting in integrated prediction results. These results are then applied to the optimization and control decisions of the electronic infusion pump system. Based on the predicted anesthesia needs, the drug flow rate and dosage for different patients are dynamically adjusted to ensure anesthesia balance for each patient. In addition, personalized settings by medical staff are combined to achieve precision anesthesia, including target depth of anesthesia and infusion schedules set by medical staff through terminal devices, thereby improving the safety of surgery and the quality of recovery for patients.

[0027] Preferably, the execution layer drive device is used to receive control commands and dynamically adjust the infusion parameters of the electronic infusion pump according to the control commands; including: adjusting the drug flow rate and dosage for different patients by controlling the execution layer device to ensure the anesthesia balance for each patient, wherein the execution layer device includes a precision stepper motor, a flow regulating valve, and a pressure sensor; wherein the precision stepper motor has an automatic calibration function, used to automatically adjust to the optimal state according to real-time flow rate and pressure data, so as to improve the control accuracy and response speed of the system; the precision stepper motor also has a self-learning function, used to optimize its own adjustment process by analyzing historical control data and current anesthesia needs, thereby achieving more efficient drug flow rate control and gradually improving control accuracy over long-term operation.

[0028] The self-learning function is based on reinforcement learning algorithms. It learns from the feedback between each adjustment operation and the anesthesia result to optimize the adjustment strategy. Especially when facing complex changes in anesthesia requirements, the self-learning function can better adapt to different working conditions and achieve refined control of the anesthesia process, thereby effectively reducing drug consumption while ensuring the quality of anesthesia.

[0029] Preferably, the process by which the execution layer drive device dynamically adjusts the infusion parameters of the electronic infusion pump according to the control instructions is as follows: The execution layer drive device first receives control instructions from the central control unit via wireless communication; the control instructions include the set target drug flow rate, dose distribution ratio, and infusion strategy, and the microprocessor built into the execution layer drive device parses the control instructions and converts them into operation signals; based on the operation signals, the execution layer drive device sends control signals to the precision stepper motors and flow regulating valves of different patients, wherein the precision stepper motors are installed at the inlet of each infusion line to automatically adjust the motor speed and the opening of the flow regulating valve according to the dose distribution ratio and target drug flow rate in the control signals, respectively, to obtain speed adjustment parameters and opening adjustment parameters, thereby changing the drug flow rate through the line; this adjustment method can very precisely control the drug infusion volume and infusion rate for each patient. The infusion rate adapts to different anesthesia needs; the pressure sensor automatically adjusts its detection sensitivity according to the infusion strategy to obtain sensitivity adjustment parameters. Based on the speed adjustment parameters, opening adjustment parameters, and sensitivity adjustment parameters, the primary infusion adjustment parameters are constructed, ensuring sufficient drug flow dynamics while avoiding unnecessary drug waste; by monitoring system pressure and flow rate, the drive unit can intelligently adjust the pump's working state to maintain stable operation of the infusion system; and through the sensing unit equipped in the execution layer drive unit, the key parameters of the electronic infusion pump are monitored in real time, including flow rate, pressure, and dosage parameters; these key parameters are fed back to the execution layer drive unit to evaluate the current anesthesia result, obtain anesthesia result feedback, and adjust the control commands based on the anesthesia result feedback, generate new control commands, and dynamically readjust the infusion parameters of the electronic infusion pump based on the new control commands to obtain secondary infusion adjustment parameters.

[0030] Preferably, the process by which the precision stepper motor automatically adjusts to its optimal state based on real-time flow rate and pressure data is as follows: Pressure sensors and flow meters installed in each infusion line of the electronic syringe pump monitor the pressure and flow rate data of the drug in real time. The pressure sensors accurately measure the real-time pressure of the drug at different locations within the pipeline, while the flow meters measure the drug flow rate through the pipeline where the precision stepper motor is located. The sensors transmit the collected pressure and flow rate data to the central control unit in the form of electrical signals, ensuring timely data acquisition and accurate transmission, providing basic data support for the adjustment of the precision stepper motor. Based on the anesthesia demand prediction model and the target anesthesia depth and infusion schedule set by the user interaction terminal device, the central control unit calculates the ideal drug flow rate and dose target values ​​for each patient or infusion line at the current moment. After receiving the real-time flow rate and pressure data, the control unit compares it with the target values, calculating the flow rate deviation and pressure deviation. The flow rate deviation is the difference between the actual flow rate and the target flow rate, and the pressure deviation is the difference between the actual pressure and the target pressure.

[0031] The precision stepper motor's built-in control algorithm is based on the PID control algorithm in classical control theory. It calculates the motor speed adjustment based on flow velocity and pressure deviations. The output control signal is calculated using three components: proportional (P), integral (I), and derivative (D) of the deviation. The proportional term reflects the magnitude of the current deviation and responds promptly to it. The integral term eliminates the system's steady-state error by accumulating the deviation over time. The derivative term predicts the system trend based on the rate of change of the deviation, suppressing overshoot. By adjusting these three parameters, the control of the precision stepper motor can be made more precise and stable. The speed adjustment signal calculated by the control algorithm is transmitted to the precision stepper motor's drive actuator: a servo motor. The servo motor controls the position of the precision stepper motor and changes its speed based on the received signal. After adjusting the speed, the precision stepper motor continues to monitor flow velocity and pressure data in real time, repeating the process of deviation calculation, control algorithm operation, and drive actuator adjustment, forming a closed-loop control system.

[0032] Preferably, the process of achieving more efficient drug flow rate control with a precision stepper motor is as follows: Feature extraction: Extract features related to drug flow rate control from historical control data and current anesthesia demand information; including: using the patient's metabolic rate change rate, the dose adjustment range set by medical staff, and the difference in flow rate changes between adjacent tubing as input features, and using the speed adjustment amount of the precision stepper motor as output features.

[0033] Model Training and Optimization: A self-learning model for the precision stepper motor is constructed using the Deep Q-Network (DQN) reinforcement learning algorithm. Extracted feature data is input into the model, which learns the optimal adjustment strategy under different conditions to achieve efficient drug flow rate control, maximizing long-term anesthesia outcomes and drug utilization efficiency. During training, model parameters are continuously optimized to accurately predict the speed adjustment actions the precision stepper motor should take under various operating conditions, achieving a more efficient drug flow rate control goal. This includes using the reward function in the DQN algorithm to evaluate the effect of each adjustment operation. If the adjusted drug flow rate is closer to the target value and drug consumption is reduced, a positive reward is given; otherwise, a negative reward is given, guiding the model to learn the optimal adjustment strategy.

[0034] Condition assessment and strategy selection: Based on real-time monitored anesthesia demand data and environmental parameters, a trained self-learning model assesses the current condition. The model selects the most suitable adjustment strategy from its learned strategy space, based on input feature data: current patient metabolic rate, dosage set by medical staff, and flow rates in other tubing, determining the direction and magnitude of the precision stepper motor's speed adjustment. Execution of adjustment: The precision stepper motor, according to the adjustment strategy determined by the model, controls the motor position through the drive actuator, achieving real-time adjustment of the drug flow rate. During adjustment, flow rate changes are continuously monitored to ensure the flow rate adjustment moves towards the target value. Simultaneously, it works in conjunction with the pressure sensor in the electronic infusion pump system to dynamically balance the flow rate distribution of each tubing according to the overall anesthesia demand of the system, avoiding local overdose or underdose of drugs and achieving efficient drug flow rate control.

[0035] Effectiveness Evaluation and Feedback Collection: Immediately after each drug flow rate adjustment, collect feedback information on the anesthesia results. This feedback includes changes in the actual depth of anesthesia, changes in drug consumption in the infusion system, and whether medical staff need to adjust the anesthesia settings again. By comparing various indicators before and after the adjustment, evaluate the effectiveness and efficiency of this adjustment in controlling the drug flow rate. If the depth of anesthesia quickly reaches the target value after adjustment and drug consumption does not increase significantly, the adjustment is considered effective. Conversely, if there are large fluctuations in the depth of anesthesia or excessive drug consumption, further analysis of the causes is required.

[0036] Model parameter update and optimization: Based on the feedback evaluation results, the parameters of the self-learning model are updated using the learning mechanism in the reinforcement learning algorithm. If a certain adjustment operation achieves a good anesthesia result and low drug consumption, the model will strengthen the parameters related to the adjustment strategy that led to the result, making it more likely to be selected in similar situations in the future. Conversely, if the adjustment effect is poor, the model will adjust the parameters accordingly to avoid using similar undesirable adjustment strategies again. By continuously repeating the above process of data collection, analysis, decision-making, feedback, and model update, the precision stepper motor can gradually improve the control accuracy during long-term operation, continuously adapt to various changes in the anesthesia system, achieve more efficient and precise drug flow rate control, and improve the stability and drug utilization efficiency of the entire anesthesia system.

[0037] Preferably, the process of achieving more efficient drug flow rate control using a precision stepper motor also includes: Historical control data accumulation: During the operation of the anesthesia system, the precision stepper motor continuously records the adjustment time, flow rate value before adjustment, flow rate value after adjustment, corresponding pressure data, patient metabolic rate at that time, surgical type data, and anesthesia parameters set by medical staff: target depth of anesthesia and infusion schedule, thus obtaining historical control data.

[0038] Current anesthesia demand monitoring: Real-time anesthesia demand information is acquired through a central control unit, including the anesthesia demand predicted for each patient based on physiological parameters such as heart rate, blood pressure, blood oxygen saturation, and individual patient characteristics, as well as anesthesia settings adjusted in real-time by medical staff through terminal devices, including temporarily increasing or decreasing the depth of anesthesia; simultaneously, the operating status of pressure sensors in the electronic infusion pump system and the speed of precision stepper motors in other tubing are closely monitored to assess their impact on current anesthesia demand. By combining the above information, the actual changes in anesthesia demand in the tubing of each precision stepper motor at the current moment are determined, thus obtaining the current anesthesia demand information.

[0039] A method for precisely controlling the anesthetic drug delivery rate using an electronic infusion pump includes the following steps: Step 1: Multidimensional data acquisition: Real-time acquisition of multidimensional data, including heart rate, blood pressure, blood oxygen saturation, drug flow rate and surgical operation status, at different monitoring points in the operating room environment and on the patient's body, and transmission of the acquired multidimensional data to the central control unit via wired or wireless communication.

[0040] Step 2: Anesthesia Demand Prediction and Analysis: Collect historical physiological data and obtain future surgical plan data, then extract anesthesia demand-related features, and use a convolutional neural network model for training to obtain an anesthesia demand prediction model; collect individual patient characteristic data, extract features to establish a patient behavior pattern model, clarify the quantitative relationship between behavior patterns and anesthesia demand, and obtain anesthesia demand prediction results based on patient behavior.

[0041] Step 3: Based on the fusion strategy, the patient behavior prediction results are integrated with the output of the prediction model of historical physiological data and surgical plan data to obtain the final anesthesia demand prediction model, predict future anesthesia demand, obtain integrated prediction results, and generate control instructions based on the integrated prediction results.

[0042] Step 4: Based on the control commands, the precision stepper motor automatically adjusts the motor speed and flow rate regulating valve opening according to the dosage distribution ratio and target drug flow rate. The pressure sensor adjusts the detection sensitivity according to the infusion strategy to achieve dynamic adjustment of drug flow rate and dosage in the infusion tubing for each patient, ensuring anesthesia balance.

[0043] Step 5: Medical staff interact with the central control unit through user interaction terminal devices to view the real-time anesthesia status and set anesthesia settings for individual patients. Attached Figure Description

[0044] Figure 1 This is a block diagram of the overall structure of the system of the present invention, showing the connection relationship between the data acquisition module, the central control unit, the execution layer driving device and the user interaction terminal device.

[0045] Figure 2 This is a schematic diagram of the data acquisition module in this invention, which includes a physiological parameter sensor, a drug flow meter, and a surgical progress monitor, and transmits data to the central control unit via communication.

[0046] Figure 3 This is a flowchart of the central control unit of the present invention, illustrating the process from receiving multi-dimensional data, processing data analysis algorithms, to generating control instructions.

[0047] Figure 4 This is a schematic diagram of the execution layer drive device of the present invention, which includes a precision stepper motor, a flow regulating valve and a pressure sensor, used to dynamically adjust the infusion parameters of anesthetic drugs.

[0048] Figure 5 This is a schematic diagram of the functional interface of the user interaction terminal device of the present invention, showing the operation interface for medical staff to view real-time anesthesia status and set anesthesia parameters for individual patients.

[0049] Figure 6 This is a flowchart of the training and prediction process of the anesthesia demand prediction model of the present invention, including the steps of historical physiological data processing, feature extraction, model training, and prediction result output.

[0050] The attached figures are labeled as follows: 1. Data acquisition module; 2. Central control unit; 3. Execution layer drive device; 4. User interaction terminal device; 5. Physiological parameter sensor; 6. Drug flow meter; 7. Surgical progress monitor; 8. Precision stepper motor; 9. Flow regulating valve; 10. Pressure sensor. Detailed Implementation

[0051] This invention provides a method and system for precisely controlling the rate of anesthetic drug delivery using an electronic infusion pump, the specific implementation of which is described in conjunction with the appendix. Figure 1 To be continued Figure 6 A detailed explanation will be provided below. The following description will introduce the function implementation, connection relationship and operation process of each module one by one to ensure that the technical solution can be fully understood and implemented.

[0052] like Figure 1As shown, the overall structure of this system includes a data acquisition module 1, a central control unit 2, an execution layer drive device 3, and a user interaction terminal device 4. These modules interact with each other via wired or wireless communication, forming a complete closed-loop control system. The data acquisition module 1 is responsible for collecting multi-dimensional data on the operating room environment and the patient's body in real time and transmitting the data to the central control unit 2 for processing. The central control unit 2 generates control commands based on the received data and sends the commands to the execution layer drive device 3 to dynamically adjust the anesthetic drug infusion parameters. The user interaction terminal device 4 provides an operating interface for medical staff to view the real-time anesthesia status and set anesthesia parameters for individual patients.

[0053] The specific components of data acquisition module 1 are as follows: Figure 2 As shown, the system includes physiological parameter sensors 5, a drug flow meter 6, and a surgical progress monitor 7. The physiological parameter sensors 5 are distributed at different monitoring points on the patient's body to collect real-time physiological parameters such as heart rate, blood pressure, and blood oxygen saturation. The drug flow meter 6 is installed in the infusion tubing of the electronic infusion pump to measure the current drug flow rate. The surgical progress monitor 7 is located in the operating room to monitor the surgical procedure status. These sensors and monitors are connected to the central control unit 2 via a communication interface to transmit the collected data in real time. To ensure the reliability of data transmission, the data acquisition module 1 adopts a redundant design, transmitting data simultaneously via both wired and wireless methods to avoid data loss due to a failure of a single communication method.

[0054] The workflow of Central Control Unit 2 is as follows: Figure 3 As shown, its core function is to receive multidimensional data from the data acquisition module 1, process the data based on a preset data analysis algorithm, and finally generate control instructions. First, the central control unit 2 acquires real-time data from the data acquisition module 1, including the patient's heart rate, blood pressure, blood oxygen saturation, drug flow rate, and surgical operation status. Then, the central control unit 2 uses data analysis algorithms to classify, clean, and normalize this data, removing outliers and noisy data to ensure the accuracy and completeness of the data. Next, the central control unit 2 analyzes the processed data based on a deep learning-based anesthesia demand prediction model, extracts features related to anesthesia demand, and trains the model using historical physiological data and surgical plan data to obtain anesthesia demand prediction results for each patient. Finally, the central control unit 2 generates control instructions based on the prediction results and sends the instructions to the execution layer drive device 3.

[0055] The structure of the execution layer drive device 3 is as follows Figure 4As shown, the system includes a precision stepper motor 8, a flow regulating valve 9, and a pressure sensor 10. The precision stepper motor 8 is installed at the inlet of each infusion line to automatically adjust its speed according to the target drug flow rate and dosage distribution ratio in the control command. The flow regulating valve 9 works in conjunction with the precision stepper motor 8 to precisely control the drug flow rate according to the opening adjustment parameters in the control command. The pressure sensor 10 is installed at a critical location in the infusion line to monitor drug pressure changes in real time and adjust the detection sensitivity according to the infusion strategy. The execution layer drive device 3 receives control commands from the central control unit 2 via wireless communication and converts them into operation signals, which are then sent to the precision stepper motor 8 and the flow regulating valve 9. The precision stepper motor 8 changes its speed according to the speed adjustment parameters in the operation signals, while the flow regulating valve 9 changes its valve opening according to the opening adjustment parameters, thereby achieving dynamic adjustment of the drug flow rate and dosage. The pressure sensor 10 monitors the system pressure in real time according to the sensitivity adjustment parameters to ensure the stability and safety of the infusion process.

[0056] The functional interface of the user interaction terminal device 4 is as follows: Figure 5 As shown, its main function is to provide medical staff with an operating platform to view real-time anesthesia status and set anesthesia parameters for individual patients. The user interaction terminal device 4 interacts with the central control unit 2 via a communication interface, displaying key parameters for each patient in real time, such as heart rate, blood pressure, blood oxygen saturation, drug flow rate, and depth of anesthesia. Medical staff can set the target depth of anesthesia and infusion schedule through the user interaction terminal device 4. These settings are transmitted as input data to the central control unit 2 to generate control instructions. In addition, the user interaction terminal device 4 also provides an alarm function; when the system detects an abnormality, it will issue an alarm prompt so that medical staff can take timely measures.

[0057] The training and prediction process of the anesthesia demand prediction model is as follows: Figure 6As shown, the core steps include historical physiological data processing, feature extraction, model training, and prediction result output. First, historical physiological data is acquired from physiological parameter sensors 5, drug flow meters 6, and surgical progress monitors 7 deployed in the operating room environment. This includes heart rate, blood pressure, blood oxygen saturation, drug flow rate, and surgical operation status at different time periods. After classification, cleaning, and normalization, this data forms a historical physiological dataset suitable for analysis and model training. Simultaneously, surgical plans and patient medical records for a future period are obtained from the hospital information system, including predicted surgical types and durations. This data is then processed using Kalman filtering to remove short-term fluctuation noise, forming a surgical plan dataset. Next, features related to anesthesia demand are extracted from the historical physiological dataset and surgical plan dataset and normalized, mapping them to the [0,1] interval to ensure different features have the same dimensions. Then, a convolutional neural network model is selected as the basic architecture for the anesthesia demand prediction model, determining the number of convolutional layers, the number of filters per layer, and the activation function. Processed historical physiological data features and corresponding historical anesthesia demand data were used as training samples and input into the constructed model for training. The Adam optimization algorithm was used to optimize and adjust the model parameters, enabling the model to learn the mapping relationship between physiological data and anesthesia demand. During training, the dataset was divided into training, validation, and test sets with a ratio of 70%:15%:15%. The model's performance metrics, Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE), were evaluated on the validation set to monitor for overfitting or underfitting. Finally, surgical plan data features were input into the anesthesia demand prediction model, and prediction results were calculated and output, including predicted anesthesia demand values ​​for different patients in the corresponding future time periods.

[0058] In practical application, the system operates as follows: Before surgery, medical staff input the patient's basic information and surgical plan through the user interaction terminal device 4, including the patient's age, weight, metabolic rate, allergy history, surgical type, and duration. The data acquisition module 1 collects the patient's heart rate, blood pressure, blood oxygen saturation, drug flow rate, and surgical operation status in real time and transmits this data to the central control unit 2. The central control unit 2 processes the received data, uses an anesthesia demand prediction model to predict the anesthesia needs of each patient, and generates control instructions. These control instructions are sent wirelessly to the execution layer drive device 3. The execution layer drive device 3 controls the precision stepper motor 8 and the flow regulating valve 9 according to the speed adjustment parameters and opening adjustment parameters in the instructions, dynamically adjusting the drug flow rate and dosage. The pressure sensor 10 monitors the system pressure in real time and adjusts parameters based on sensitivity to ensure the stability of the infusion process. Throughout the surgery, medical staff can view the anesthesia status of each patient in real time through the user interaction terminal device 4 and adjust the anesthesia parameters according to the actual situation. The precision stepper motor 8 in the execution layer drive device 3 has a self-learning function. By learning from historical control data and current anesthesia needs, it continuously optimizes the adjustment strategy and gradually improves the control accuracy. At the same time, the pressure sensor 10 in the execution layer drive device 3 works in conjunction with the precision stepper motor 8 to dynamically balance the flow rate distribution of each tubing according to the overall anesthesia needs of the system, avoiding local drug overdose or underdose, and achieving efficient drug flow rate control.

[0059] As can be seen from the above embodiments, the present invention achieves precise control of the anesthetic drug infusion process through the coordinated work of the data acquisition module 1, the central control unit 2, the execution layer driving device 3, and the user interaction terminal device 4. The clear connection relationships and data interaction mechanisms between the modules ensure the efficient operation and stability of the system.

[0060] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principle of this invention will be further explained below in conjunction with a specific application scenario.

[0061] Before the surgery begins, medical staff input the patient's basic information and surgical plan through the user interaction terminal device 4. This information includes the patient's age, weight, metabolic rate, allergy history, as well as the type and duration of the surgery. The data acquisition module 1 then activates, with its internal physiological parameter sensors 5 distributed at different monitoring points on the patient's body to collect physiological parameters such as heart rate, blood pressure, and blood oxygen saturation in real time; a drug flow meter 6 is installed in the infusion tubing of the electronic infusion pump to measure the current drug flow rate; and a surgical progress monitor 7 is placed in the operating room to monitor the surgical procedure status. The sensors and monitors transmit the collected data to the central control unit 2 via wired or wireless communication. To avoid data loss due to a failure of a single communication method, the data acquisition module 1 adopts a redundant design, that is, it transmits data simultaneously through both wired and wireless methods.

[0062] After receiving the multidimensional data, the central control unit 2 first classifies, cleans, and normalizes it to remove outliers and noisy data, ensuring the accuracy and completeness of the data. Subsequently, a deep learning-based anesthesia demand prediction model analyzes the processed data. This model extracts features related to anesthesia demand from historical physiological datasets and trains them in conjunction with surgical planning datasets. Specifically, historical physiological data includes heart rate, blood pressure, blood oxygen saturation, drug flow rate, and surgical procedure status at different time periods, while surgical planning data includes surgical type and duration information. This data is processed using Kalman filtering to remove short-term fluctuation noise, forming a dataset with a unified format. A convolutional neural network model serves as the basic architecture, and the Adam optimization algorithm is used to adjust the model parameters, enabling the model to learn the mapping relationship between physiological data and anesthesia demand. Finally, the central control unit 2 generates control instructions based on the prediction results and sends them to the execution layer driver 3.

[0063] The precision stepper motor 8 and flow regulating valve 9 in the execution layer drive unit 3 receive control commands from the central control unit 2. The precision stepper motor 8 is installed at the inlet of each infusion line, automatically adjusting its speed according to the target drug flow rate and dosage distribution ratio specified in the command. The flow regulating valve 9 precisely controls the drug flow rate by adjusting parameters according to the opening degree specified in the command. Pressure sensors 10 are installed at key locations in the infusion lines to monitor drug pressure changes in real time and adjust parameters based on sensitivity to ensure the stability of the infusion process. Specifically, the precision stepper motor 8 incorporates a PID control algorithm, calculating the motor speed adjustment based on flow rate and pressure deviations. The proportional term reflects the current deviation magnitude, the integral term eliminates steady-state error, and the derivative term predicts system trends, thus achieving closed-loop control. Furthermore, the precision stepper motor 8 has a self-learning function, continuously optimizing the adjustment strategy through the DQN model in the reinforcement learning algorithm to adapt to complex changes in anesthesia requirements.

[0064] Throughout the surgery, medical staff can monitor the anesthesia status of each patient in real time through the user interaction terminal device 4. The user interaction terminal device 4 interacts with the central control unit 2 via a communication interface, displaying key parameters such as heart rate, blood pressure, blood oxygen saturation, drug flow rate, and depth of anesthesia for each patient in real time. When medical staff detect abnormalities, they can set the target depth of anesthesia and infusion schedule through the terminal device. These settings are transmitted as input data to the central control unit 2 to generate new control instructions. The execution layer drive device 3 dynamically adjusts the drug flow rate and dosage according to the new instructions to ensure balanced anesthesia for each patient. Simultaneously, the pressure sensor 10 and the precision stepper motor 8 work together to dynamically balance the flow rate distribution of each tubing according to the overall anesthesia needs of the system, avoiding local overdose or underdose of drugs.

[0065] As can be seen from the above steps, this invention achieves precise control of the anesthetic drug infusion process through the collaborative work of the data acquisition module 1, the central control unit 2, the execution layer drive device 3, and the user interaction terminal device 4. The data acquisition module 1 is responsible for real-time acquisition of multi-dimensional data; the central control unit 2 generates control commands through data processing and analysis; the execution layer drive device 3 dynamically adjusts the drug flow rate and dosage according to the commands; and the user interaction terminal device 4 provides an operating platform for medical personnel. The clear connection relationships and data interaction mechanisms between the modules ensure the efficient operation and stability of the system.

[0066] All content not described in detail in this specification is prior art known to those skilled in the art, and the model parameters of each electrical appliance are not specifically limited; conventional equipment can be used. Electrical control components not mentioned in this technical solution are prior art and are therefore not shown in the figures, nor will they be described further here.

[0067] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for precise control of the drug injection rate of an electronic syringe pump, characterized in that, The method comprises the following steps: Step 1: Multi-dimensional data acquisition: Real-time multi-dimensional data including patient heart rate, blood pressure, blood oxygen saturation, current drug flow rate and operation state are collected by data acquisition modules (1) distributed in the operating room environment and different monitoring points of the patient's body, and are transmitted to the central control unit (2) through wired or wireless communication; Step 2: Anesthesia demand prediction and analysis: Historical physiological data and future operation plan data are collected, anesthesia demand related features are extracted, a convolutional neural network model is selected for training to obtain an anesthesia demand prediction model; patient individual feature data are collected, features are extracted to establish a patient behavior pattern model, the quantitative relationship between the behavior pattern and the anesthesia demand is determined, and anesthesia demand prediction results based on patient behavior are obtained; Step 3: The patient behavior prediction results, historical physiological data and operation plan data prediction model outputs are integrated based on a fusion strategy to obtain a final anesthesia demand prediction model, future anesthesia demand is predicted, integrated prediction results are obtained, and control instructions are generated based on the integrated prediction results; Step 4: Based on the control instructions, the precision stepper motor (8) automatically adjusts the motor speed and the opening degree of the flow regulating valve (9) according to the dose distribution ratio and the target drug flow rate, and the pressure sensor (10) adjusts the detection sensitivity according to the infusion strategy to realize dynamic adjustment of the drug flow rate and dose of each patient's infusion pipeline, and ensure anesthesia balance; Step 5: Medical staff interact with the central control unit (2) through the user interaction terminal device (4), view real-time anesthesia state and set individual patient anesthesia.

2. The method of claim 1, wherein the method comprises: The central control unit (2) receives multi-dimensional data, processes the collected multi-dimensional data based on a preset data analysis algorithm, identifies the anesthesia demand differences of different patient individuals, and obtains anesthesia demands of different patients. The data analysis algorithm comprises an anesthesia demand prediction model based on deep learning, which is used to predict anesthesia demands of different patients in a future period of time according to historical physiological data, operation type and external environmental variables.

3. The method of claim 2, wherein the method further comprises: determining the volume of the drug to be injected; and determining the volume of the drug to be injected based on the volume of the drug to be injected and the volume of the drug to be injected. The anesthesia demand prediction model predicts anesthesia demands of different patients in a future period of time as follows: The physiological parameter sensors (5), drug flow meters (6) and operation progress monitors (7) arranged in the operating room environment obtain historical physiological data including heart rate, blood pressure, blood oxygen saturation, drug flow rate and operation state in different time periods; and the historical physiological data are classified, cleaned, and abnormal values and error data are removed to ensure the accuracy and integrity of the data; The sorted historical physiological data are stored according to patient individuals and time sequences to form a historical physiological data set for analysis and model training; The data receiving unit of the central control unit (2) obtains operation plans and patient medical record data in a future period of time from a hospital information system, including predicted operation type and operation duration information; And the Kalman filtering method is used to remove short-term fluctuation noise in the data to obtain an operation plan data set; The features related to anesthesia demand are extracted from the historical physiological data set and the surgery plan data set, the related features are obtained, and normalization processing is performed to map them to [0, 1], thereby obtaining the historical physiological data features and the surgery plan data features.

4. The method of claim 3, wherein the method further comprises: determining the volume of the drug to be injected; and determining the volume of the drug to be injected based on the volume of the drug to be injected and the volume of the drug to be injected. The basic architecture of the anesthesia demand prediction model is selected as a convolutional neural network model, the number of convolutional layers, the number of filters in each layer, and the activation function are determined; including: setting a neural network containing four convolutional layers, the number of filters in each layer is [128, 64, 32, 16], and the activation function adopts the Leaky ReLU modified linear unit function; The processed historical physiological data features and the corresponding historical anesthesia demand data are used as training samples, and are input into the constructed model for training, and the Adam optimization algorithm is used to optimize and adjust the model parameters, so that the model learns the mapping relationship between physiological data and anesthesia demand.

5. The method of claim 4, wherein the method further comprises: determining the volume of the drug to be injected; and determining the volume of the drug to be injected based on the volume of the drug to be injected and the volume of the drug to be injected. In the training process, the data set is divided into a training set, a validation set and a test set, and the proportion is set to 70%:15%:15%; the performance indicators root mean square error RMSE and mean absolute percentage error MAPE of the model are evaluated on the validation set to monitor whether the model has overfitting or underfitting phenomenon; if the performance indicators on the validation set no longer improve or start to deteriorate, the training is stopped, and the anesthesia demand prediction model is obtained.

6. The method of claim 1, wherein the method further comprises: determining a target flow rate of the anesthetic; and adjusting the flow rate of the anesthetic to the target flow rate. The anesthesia demand prediction is performed based on the patient behavior pattern combined with the individual feature data of the patient, and the process is as follows: The individual feature information actively input by the patient is collected through the user interaction terminal device (4), and the individual feature information includes the patient's age, weight, metabolic rate, allergy history, and whether there is a specific anesthesia tolerance rule; Indirect data generated by the patient interacting with the anesthesia system are collected, and individual feature data are analyzed; Potential individual feature data are mined from the patient's past medical record information; The collected patient individual feature data are cleaned to remove invalid, incorrect or duplicate data records; and data of different formats and types are converted to a unified format; relevant features are extracted according to the characteristics of the patient individual feature data and the prediction results; and composite features are constructed combined with physiological data and time information.

7. The method of claim 6, wherein the method further comprises: determining a target flow rate of the anesthetic; and adjusting the flow rate of the anesthetic to the target flow rate. Patients with similar individual features are classified into the same class through cluster analysis, and a typical behavior pattern model is established for each cluster; The anesthesia response pattern of the patient at different time points is found by using the association rule mining algorithm, the frequency and conditions of its occurrence are analyzed, and the dynamic anesthesia demand change rule of the patient under different conditions is determined; the average anesthesia demand and anesthesia demand fluctuation range under different behavior patterns are calculated by statistical analysis method, and the mapping relationship model between behavior pattern and anesthesia demand is established.

8. The method of claim 7, wherein the method further comprises: determining a target flow rate of the anesthetic; and adjusting the flow rate of the anesthetic to the target flow rate. The process of dynamically adjusting the infusion parameters of the electronic injection pump by the execution layer driving device (3) according to the control instruction is as follows: The execution layer driving device (3) first receives the control instruction from the central control unit (2) through wireless communication; the control instruction includes the set target drug flow rate, dose allocation ratio and infusion strategy; The operation signal based layer driving device (3) sends control signals to the precision stepper motor (8) and the flow regulating valve (9) of different patients, wherein the precision stepper motor (8) is installed at the inlet of each infusion pipeline, and is used for automatically adjusting the motor speed and the opening degree of the flow regulating valve (9) according to the dose distribution ratio and the target drug flow rate in the control signal respectively, so as to obtain the speed adjustment parameter and the opening degree adjustment parameter respectively.

9. The method of claim 8, wherein the method further comprises: determining the volume of the drug to be injected; and determining the volume of the drug to be injected based on the volume of the drug to be injected and the volume of the drug to be injected. The process of automatically adjusting the precision stepper motor (8) to the optimal state according to the real-time flow rate and pressure data is as follows: The pressure and flow rate data of the drug are monitored in real time through the pressure sensor (10) and the flowmeter installed in each infusion pipeline of the electronic injection pump. According to the anesthesia demand prediction model and the target anesthesia depth and infusion time table set by the user interactive terminal device (4), the central control unit (2) calculates the ideal drug flow rate and the dose target value corresponding to the current time for each patient or infusion pipeline. After the control unit receives the real-time flow rate and pressure data, it compares the data with the target value, calculates the flow rate deviation and the pressure deviation, and the flow rate deviation is the difference between the actual flow rate and the target flow rate, and the pressure deviation is the difference between the actual pressure and the target pressure.

10. The method of claim 9, wherein the method further comprises: determining a target flow rate of the anesthetic; and adjusting the flow rate of the anesthetic to the target flow rate. The control algorithm built in the precision stepper motor (8) is based on the PID control algorithm in the classical control theory, calculates the speed adjustment amount of the motor according to the flow rate deviation and the pressure deviation, calculates the output control signal according to the proportion P, the integral I and the differential D of the deviation, and the speed adjustment signal calculated by the control algorithm is transmitted to the servo motor of the driving execution mechanism of the precision stepper motor (8); the servo motor controls the position of the precision stepper motor (8) according to the received signal, and changes the speed of the motor.

Images